LTX 2.3 already has a baseline understanding of scene changes, however I have noticed that it can be difficult to properly anchor sequentially. This model adds a trigger that can be used to more easily anchor the transition event. The trigger phrase is "immediate scene transition," which can be inserted into the prompt such as the following:
<initial scene>. Then an immediate scene transition to <new scene>. After the scene transition, <new scene events>.It is not always perfect in preserving subject details, however this can be compensated by specifying that an element remains unchanged, for example, this prompt will change the outfit but preserve the hairstyle:
immediate scene transition to the same person inside a a home and sitting at the kitchen table and sipping a cup of coffee, she has the same hairstyle and is wearing jeans and a white tshirt.California AB 2013 Training Data Disclosure
This LoRA was fine-tuned using visual data consisting of synthetic clips. The training data may include copyrighted material owned by third parties. No training data was licensed or purchased. This LoRA is provided for non-commercial use only under the terms of its distribution.
The dataset consists of over 50 clips. This is the total used for training after low quality sets were culled. Dataset was collected/created in 2026.
Data was processed through standard resizing, cropping, normalization, and labeling steps. Synthetic entries were included as part of the training dataset.
This model is intended for non-commercial, experimental, and educational use. Generated outputs may reflect copyrighted visual styles or themes associated with the underlying training data. Users are responsible for ensuring compliance with applicable copyright law, other intellectual property laws, and all other applicable laws.
Description
FAQ
Comments (10)
thank you from the bottom of my heart! <3 was gonna make an updated version of mine but laziness got the better of me
I’m curious to try out this LORA to see how it might influence the character to follow. Thanks for sharing :)
Here’s my video—there’s a small glitch in the transition around the genital area (“penis”), but actually, it’s not bad—I’m surprised. Excellent work! Now I just need to figure out the trick, come up with other shots, and get creative. Thanks again :)
Super work, works really well.
Thanks so much for this lora. Yesterday, I was cursing the lack of a quick transition lora for LTX2.3, casually checked Civitai and found this! Initial tests seem to be working very well.
Amazing work. You've been able to pull out more than I expected from LTX. But, LTX sure makes it rough on transitions... It's pretty hard to match Wan. Still, this is extremely useful.
well done, works great. Just curious how do you train for a scene change please?
I made synthetic clips using WAN 2.2. I used Klein to create a matching subject in two different scenes, animated them with WAN, then combined both videos into one. Then when captioning them, I followed a formula consistent with the model's trigger.
seems better for hard cuts, while the other transition lora is better for slow cuts.
Hi, very nice lora, thank you. I was testing it and it seems best for 1 subject and 2 takes. If its possible, would be nice if you made one for 3 takes. I could make it do 3 takes but it is obvious that it struggles at them.
Details
Files
LTX2.3_SceneChange.safetensors
Mirrors
Immediate Scene Transition - LTX2.3 - immediate scene transition,initial scene.then an immediate scene transition to new scene.after the scene transition,new scene events.safetensors
Immediate Scene Transition - LTX2.3 - immediate scene transition,initial scene.then an immediate scene transition to new scene.after the scene transition,new scene events.safetensors
207-SceneChange.safetensors